Rich 1

Controlled Spalling In Lithium Niobate: A Universal Mechanical

Release Technology For Inorganic Thin-Film Devices

D. Rich1 and C. Bayram1

1Innovative Compound Semiconductor Laboratory, Department of Electrical and Computer Engineering,

University of Illinois at Urbana-Champaign

I. Abstract

Scalable and cheap fabrication of thin films (less than 50 μm thick) has the potential to

revolutionize aerospace, bioengineering, and consumer electronics with small and flexible

electronic devices. Controlled spalling, a method capable of producing these benefits and more,

controls the stresses within a substrate and allows for large-scale release of the desired layer. In

this work, controlled spalling is implemented in a laboratory setting using electroplating to produce

these stresses, then analyzed and refined while applying it to the case of thin-film lithium niobite

transducers.

II. Introduction

Thin-film lithium niobate (LiNbO3) structures have been shown to be extremely useful as surface

acoustic wave (SAW) transducers [1]. These transducers are useful as mobile communication

bandpass devices, and, in particular, transducers operating from 3-30 GHz, the so-called super

high frequency (SHF) range, allowing for faster data transfer rates [2] as well as greater sensor

selectivity [3]. LiNbO3 thin films have been demonstrated to have high coupling coefficients (K2)

and stability at frequencies in the SHF range due to the material’s piezoelectric properties [4].

However, to achieve these performance metrics, conductive interdigital transducers must be

fabricated on the LiNbO3 surface. In this work, we concern ourselves with the fabrication of these

thin-film transducers. A method of creating thin-film LiNbO3 metal transducer devices easily and

cheaply would lead to more accessible technology for a wide variety of communications and

MEMS applications.

LiNbO3 single-crystal thin films can be deposited by chemical vapor deposition on other

crystalline materials, most commonly silicon [5]. However, production of thin-films in this manner

requires additional deposition and etching. Etching of LiNbO3 is a volatile process that can lead to

contamination of other samples, and is best avoided. We propose a better approach: to release a

thin film from a bulk substrate after fabrication of that device.

In the past several decades, various methods for brittle thin-film release have been explored.

Physical release methods such as epitaxal lift-off [6], which selectively etches a substrate from a

thin film, limit the variety of materials that can be used and can damage devices if the etchant is

not compatible. Optical methods, meanwhile, include release methods such as laser lift-off [7],

which aims a tuned laser at a film-substrate interface which is excited to release upon light

absorption. Optical methods are expensive and limited in scalability due to the small-area beam

shape. I address the need for a universal mechanical release method, opening the door to LiNbO3

thin-film devices at low cost and accurate thickness.

Rich 2

III. Dynamics of Controlled Spalling

Controlled spalling is understood by separating the stress at the crack tip into directional modes,

each with a well-defined role [8, 9], as shown in Fig. 1. Mode I stress, the component of the stress

perpendicular to the tensile layer, acts to open the crack. Mode II stress, meanwhile, is the shear

stress acting directly at the crack tip, parallel to its propagation.

The magnitudes of both stresses, denoted KI and KII, are calculated according to Eqns. (1) and (2)

[8].

�� ��

√2�ℎcos �

�

√2�ℎ�sin � � ��

��� ��

√2�ℎsin �

�

√2�ℎ�cos � � ��

Suo et. al. [8] finds that cracks deflect in a direction that

decreases KII as shown in Fig. 2. The result of any crack

propagation, therefore, will be a unique, constant depth of

cracking (and thin-film thickness) at which mode II stress

is eliminated.

a. Application to LiNbO3

LiNbO3 is brittle, so this theory applies. If one can find a

way to manipulate the stresses in a pre-fabricated LiNbO3

device, a thin film can be created with excellent control of

thickness based on the calculation of KII stress. It is

additionally convenient in that the required contacts for a

SAW transducer can be fabricated on the substrate before

creation of the thin film, greatly easing the process.

IV. Overview

When the summer began, I had already designed a

methodology for the spalling of silicon wafers, to cheaply

test all components of the process before expanding to

LiNbO3. The process was not yet consistent, but after the

work of this summer that has begun to change. It consists of three major steps: the deposition of

the stressed layer so that spalling can be made possible, the characterization of the system to ensure

everything will occur as expected, and spalling itself, where the thin film is peeled up using the

stresses imparted by the deposited layer. This summer, these steps were tackled systematically and

in order so that consistent, predictable results could be achieved.

a. Stressed Layer Deposition

I chose electroplating, the transfer of Ni+ ions from solution to a negatively biased substrate, to

meet the significant challenge of controlling the thickness and stress of the final tensile layer. I

(2)

(1)

KII = 0

KII < 0

KII > 0

Figure 1. In (a), a demonstration

of the spalling action, with a

substrate-based pre-fabricated

device in place [5]. In (b), the

effects of KII on deflection of the

crack tip [5]. Blue arrows are the

stresses required to produce that

deflection.

a)

b)

Rich 3

carefully designed a 3D-printed, substrate-specific holder to take into account the variable

curvature of the sample due to its internal stress. The result allowed for easy mounting and minimal

handling while maximizing conductive contact with the sample. However, direct electroplating on

Si substrates causes failure almost immediately.

The challenge of adhesion of the tensile layer to the substrate was met in two ways. First, a thin

Ni ‘seed’ layer was deposited with electron beam evaporation, a high-adhesion method in which a

gas of the deposition atom contacts the substrate. Second, adhesion was enhanced with a layer of

titanium, which bonds with both the substrate and Ni, deposited between the two materials.

Over the summer, however, I found that the deposition of this

seed layer was very inconsistent. I narrowed this down to two

problems: first, the deposition system had poor beam control,

causing inconsistent layer quality, and second, samples tended to

break when removed from the double-sided tape that allowed

them to hang upside-down in the deposition chamber. I solved

these problems, and rendered deposition consistent, by switching

to a new system with better control and developing a method of

taping that was easy to remove (shown in Fig. 2). The darker

portions, which aren’t sticky on the bottom, can be cut along the

red lines so that the sample can be lifted from the surface and

peeled up. This method is original and applicable to any

breakable sample.

I also noticed that, in some cases, the seed layer tended to peel up. I attributed this to cleanliness:

particles that get between the layers and interfere with adhesion. After implementing piranha

cleaning before deposition, this problem disappeared, leaving seed layer deposition consistent.

Electroplating, then, needed to be refined. Many samples were

electroplated, and the problem of electrical contact became

apparent. A holder was designed and 3D-printed that uses o-ring

seals to prevent nickel deposition on a metal ring that holds the

sample down and provides that contact. With this, electroplating

area was decreased, but all failures from the edge were

eliminated and plating became quite consistent.

After this, Nomarski microscopy was used to detect defects in

the film, and many were found. After testing several theories, it

was found that rinsing the samples with water after deposition

would prevent the electroplating solution from remaining on the

wafer and dissolving the nickel layer. Implementing this rinsing

resulted in flawless, consistent samples, allowing me to move on

to the analysis step.

Figure 2. ‘Band-aid’

method of sample taping

Figure 3. One of many

high-quality stressed Si

samples

Rich 4

b. Stressed Sample Analysis

Two methods of sample analysis after plating were developed. The first of these is designed to

analyze the sample’s thickness uniformity. The nickel is selectively etched away so that a step

height from the top of the electroplated layer to the silicon below can be measured using a

profilometer at many different places through the center of the wafer. A wafer so etched is shown

in Fig. 4a, and the results of this etching at many different times of plating are shown in Fig. 4b.

Figure 4. Results of refined etching and thickness uniformity analysis

Results suggest excellent thickness uniformity except around the edges. This ‘drop-off’ area is

highlighted in Fig. 4a, and is almost certainly due to electroplating inhibition by the o-ring. It is

yet to be seen if this non-uniformity will affect spalling.

The second method of analysis optically

measures the curvature of the wafer surface, then

uses Stoney’s equation to calculate stress [10].

Although the machine used to perform this

measurement is difficult to understand, it was

verified through other methods and I believe it is

well-understood. The results of stress

measurements, as well as averaged thickness

measurements from etching, on different plated

samples is shown in Fig. 5.

V. Conclusion and Future Work

Although this research is not yet complete, results

are very promising. Consistent spalling, along

with stress and uniformity results that align well

with theory, are indicative of a process that is robust and well-suited to incremental adjustment to

achieve controlled spalling in LiNbO3.

a)

b)

Figure 5. Stresses and thicknesses at

different electroplating times